AI Coding Is Exposing Who Actually Understands Their Product

There's a conversation I keep having with CTOs. It usually starts with someone worrying about the "AI slop factory," the idea that because AI lets you ship so many features, all you'll do is clutter up your product with stuff that doesn't move the needle. Overwhelm your UI. Drown your users in features nobody asked for.

And look, it's a real concern. But the diagnosis is wrong.

The slop factory was always running. It was just a slow slop factory. Companies have always shipped products that didn't deliver value. Features that looked good on a roadmap and died quietly in production. The difference is that when it took a team three months to build something useless, you could blame the timeline. "We didn't get enough shots on goal." "Engineering was the bottleneck." "We need to move faster."

Now you can move faster. And the features are still useless.

The Mask Has Come Off

Speed was never the problem. It was a mask. A very convincing one, because it felt true. When engineering is slow, you never get to find out whether your ideas were any good. You ship two things a quarter, neither moves the metrics, and you tell yourself it's because you didn't have enough throughput to find the winners.

AI has removed that excuse. You can now build and ship in days what used to take months. If you dream it, you can build it. So what happens when companies start shipping ten times as many features and the metrics still don't move?

What's exposed is something uncomfortable: a bankruptcy of product thinking.

I've Seen This Film Before

I had a front-row seat to exactly this dynamic, years before AI coding tools existed.

When I joined Airtasker as VP of Engineering, the company-wide sentiment was clear: engineering was the problem. And honestly, it was partly true. The engineering team hadn't kept up with the scale of the business. Technical debt had compounded to the point where shipping anything meaningful was painfully slow.

The product and design teams had a comfortable position. They had this beautiful backlog of things they wanted built. Brilliant ideas, waiting in line. If only engineering could deliver them.

So I got to work. With the help of a lot of fantastic colleagues, we rebuilt the engineering team's ability to deliver. We got to a point where we could ship product at a reasonable pace. Good. Problem solved, right?

Not quite. We started shipping, and the business metrics didn't improve. The features we were being asked to build, the ones that had been sitting in that beautiful backlog, weren't the right features. They didn't solve the problems they were supposed to solve.

The slow engineering team had been masking the real issue all along. When the mask came off, what was underneath wasn't pretty. That experience is a big part of why I ended up moving into the COO role. I started caring about how all of these pieces — engineering, product, design, growth — needed to work together. Because fixing one without fixing the others just moves the bottleneck.

The Bottleneck Has Moved

This is what's happening across the industry right now, at scale.

AI coding tools have moved the bottleneck. For years, product and design could operate at whatever quality level they liked, because engineering throughput was the constraint. There was always a ready-made excuse for why things weren't working. We just need to build more. We just need to ship faster. We just need more engineers.

Now building is nearly free. And the bottleneck is sitting squarely on the question: do you actually know what to build?

Most teams, it turns out, don't. Not really. They have backlogs full of features, sure. They have roadmaps and quarterly plans and OKRs. But the hard work of understanding your market, defining the problem precisely, designing solutions that actually fit how your users think and behave? That's rare. That was always rare. It was just harder to notice when you could only test two ideas a quarter.

Design Is the Sleeper Skill

If I had to bet on which discipline becomes the most valuable in this new era, I'd bet on design. Not visual design specifically (though that matters), but the broader practice of understanding humans and building the right thing for them.

Product management matters too, obviously. But product management without strong design instincts tends to produce feature lists rather than coherent experiences. And feature lists are exactly what the slop factory runs on.

The companies that are going to win are the ones that can answer a deceptively simple question: what problem are we solving, and what is the best way to solve it? That question requires deep understanding of your market. It requires research. Empathy. Taste. Things that don't get easier just because you can build faster.

Same as Ever

Morgan Housel wrote a book called "Same as Ever." It's about finance, mostly, but the core idea is useful here: when everything around you is changing, focus on what isn't.

What's same as ever is that knowing the right thing to build is hard. It has always been hard. It was hard before AI, it was hard before agile, it was hard before the internet. Understanding a market, identifying real problems, designing solutions that people actually want to use. That is genuinely difficult work, and no amount of engineering speed makes it easier.

Every other bottleneck in software is dissolving. This one isn't going anywhere.